Metraipolis › archive › dc_c183ec8a6170

VoxLock: Target-Speaker Extraction for Voice Agents Without an Enrollment Step — arXiv:2610.40727

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.40727
License (as archived)
CC0-1.0
Archived
2026-11-03T06:00:52.159Z
SHA-256 of the archived text
0577122bf57ae85e38d4c25ec0cfa563f55be27a75efdd4a1d4d96d25c1c3367
Stable URL of this page
https://metraipolis.com/documento/dc_c183ec8a6170

How to cite

VoxLock: Target-Speaker Extraction for Voice Agents Without an Enrollment Step — arXiv:2610.40727 — arXiv (en), https://arxiv.org/abs/2610.40727 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c183ec8a6170 (Metraipolis archive dc_c183ec8a6170, sha256 0577122bf57ae85e38d4c25ec0cfa563f55be27a75efdd4a1d4d96d25c1c3367)
BibTeX
@misc{dcc183ec8a6170,
  title = {VoxLock: Target-Speaker Extraction for Voice Agents Without an Enrollment Step — arXiv:2610.40727},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.40727},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c183ec8a6170; sha256 0577122bf57ae85e38d4c25ec0cfa563f55be27a75efdd4a1d4d96d25c1c3367},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.40727. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

A voice agent's speech recognizer transcribes everyone near the microphone, including people other than the caller. I fine-tuned a public causal target-speaker extraction model (REAL-TSE, 27M parameters) on one GPU to work from a voice sample cut from the call itself, and released it as VoxLock. I score it by what a recognizer writes on real recordings. On Krisp's 265-recording benchmark it reaches 4.95% WER against 23.21% unprocessed and lets 9 background words into the transcript over 85 minutes of background-only speech. Its WER is not significantly different from Krisp's best filters and is significantly lower than the other commercial and open systems I tested. On 400 AMI meeting windows built and frozen before scoring, it cuts WER from 42.95% to 10.63%. The 8-bit model runs a live call on 77% of one CPU thread with 24 ms of algorithmic latency. Model, code and per-file scores are open.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv eess.AS, nuovi lavori del giorno).
Categorie arXiv: Audio and Speech Processing (eess.AS).
Pagina del lavoro: https://arxiv.org/abs/2610.40727 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-03T05:42:49.539Z dalla pagina https://arxiv.org/list/eess.AS/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).